Chapter 1

THE EQUATIONS OF
FLUID
DYNAMICSDRAFT
The equations of fluid mechanics are derived from first principles here, in order
to point out clearly all the underlying assumptions. The equations can take
various different forms and in numerical work we will find that it often makes a
difference what form we use for a particular problem.
We will work solely with the continuum theory of fluids, and thus use conservation principles, supplemented by constitutive assumptions about the nature
of the fluids. The conservation principles are common to any material where
the continuum hypothesis is valid but different constitutive hypothesizes apply
to different materials. Expressing the basic principles of conservation of mass,
momentum, and energy in mathematical form leads to the governing equations
for fluid flow.
Here we derive the equations for fluid motion, with particular emphasize on
incompressible flows.

1.1

General Flows

The principle of conservation of mass states that mass can not be created nor
destroyed. Therefore, if we consider a volume fixed in space, V , then the change
of mass inside this volume can only take place if mass flows in or out through
the boundary of this volume, S.1 Stated more precisely
Z
I
d
dv = u nds,
(1)
dt V
S
1 In standard text books the fundamental laws are often stated for a volume of fluid moving
with the fluid. In computational work the elementary volumes are usually stationary, therefore
it is simpler to start with a stationary volume.

CHAPTER 1. THE EQUATIONS OF FLUID DYNAMICSDRAFT

where n is the outward normal, the density and u the velocity. Here, the left
hand side is the rate of change of mass in the volume V and the right hand side
represents in and out flow through the boundaries of V . Since the volume is
fixed in space we can take
R the derivative
H inside the integral, and by applying
the divergence theorem ( V adv = S a nds) to the boundary fluxes on the
right hand side we have

Z 

+ (u) dv = 0.
(2)
t
V
This must hold for any arbitrary volume, no matter how small, and thus must
also hold at a point, if the flow field is smooth. The partial differential equation
expressing conservation of mass is therefore:

+ (u) = 0
t

(3)

By using the definition of the substantial derivative

()
D()
=
+ u (),
Dt
t

(4)

the continuity equation can be rewritten as

D
+ u + u =
+ u = 0.
t
Dt

(5)

The equation of motion is derived by assuming that Newtons 2nd law of

motion is valid for any arbitrary volume cut out of the fluid. Thus, the rate of
change of momentum of a fixed volume is the net momentum flux across the
boundaries of the volume plus the net forces acting on the volume. Therefore
Z
I
Z
I
d
udv = u(u n)ds +
f dv +
nTds.
(6)
dt V
S
V
S
The first term on the right hand side is the momentum flux through the boundary of V , and the next terms are the body force f and surface forces, respectively.
T is a symmetric stress tensor constructed in such a way that nT are the forces
on the surface of V .
By the same arguments as applied to the mass conservation equation this
must be valid at every point in the fluid, so
u
= (uu) + f + T.
t

(7)

Here, uu is a tensor whose ij-th component is ui uj . The nonlinear term can be

written as
(uu) = u u + u (u).
(8)
Using the definition of the substantial derivative and the continuity equation
we can rewrite the above equation as

Du
= f + T.
Dt

(9)

1.1. GENERAL FLOWS

This is Cauchys equation of motion and is valid for any continuous medium. For
fluids like water, oil, and air (as well as many others that are generally referred
to as Newtonian fluids) the stress may be assumed to be a linear function of the
rate of strain, or
T = (p + u)I + 2D
(10)
where I is the unit tensor and D = 12 (u + uT ) is the deformation tensor
whose components are
1  ui
uj 
Dij =
+
.
(11)
2 xj
xi
Here, is the second coefficient of viscosity, and if Stokes hypothesis is assumed
to hold then = (2/3). 2 For incompressible flows, the divergence of the
velocity field u, is identically zero, so any questions about the validity of
Stokes hypothesis are irrelevant. Substituting the expression for the stress tensor
into Cauchys equation of motion results in

Du
= f p + ( u) + 2D.
Dt

(12)

Conservation of energy leads to

Z
I
Z
I
d
1 2
1 2
(e+ u )dv = (e+ u )unds+ uf dv+ n(uTq)ds (13)
dt V
2
2
S
V
S
where u2 = u u. The term on the left hand side is the sum of the internal and
kinetic energy, the first term on the right hand side is flow of internal and kinetic
energy across the boundaries, the second term represents work done by body
forces, the third term is the work done by the stresses at the boundary (pressure
and viscous) and q is the heat-flux vector. This equation can be simplified by
using the momentum equation: Taking the dot product of the velocity with
equation (9) gives

u2 /2
= u u2 /2 + u f + u ( T)
t

(14)

for the mechanical energy. After using this equation to cancel terms in (13) and
applying the same arguments as before we obtain the convective form of the
energy equation
De

T u + q = 0.
(15)
Dt
This equation needs to be supplemented by constitutive equations for the specific
fluids we are considering.
If we use the assumptions for the stress tensor made earlier (Newtonian fluids), and if we also assume that the flux of heat is proportional to the gradient of
the temperature, T (this is called Fouries law and k is the thermal conductivity)
q = kT,
2 In

(16)

several texts the discussion is based on the bulk viscosity = 2/3 + . Stokes
hypothesis is that the bulk viscosity vanishes.

CHAPTER 1. THE EQUATIONS OF FLUID DYNAMICSDRAFT

and radiative heat transfer is negligible, then the energy equation takes the form

De
+ p u = + kT
Dt

(17)

Here, = ( u)2 + 2D D is called the dissipation function. It can be

shown that , which represents the rate at which work is converted into heat,
is always greater or equal to zero. In general we also need an equation of state
giving, say, pressure and temperature as a function of density and energy and
equations for , and k.
p = p(e, ),

T = T (e, )

(18)

as well as equations for , and k.

1.2

Special Cases

In many practical applications, considerable savings can be achived by working

with simplified form of the governing equations. Thus, the flows can be assumed
to be invisicid if boundary layers do not play a large role, and incompressible if
sound waves are not important.

1.2.1

Compressible, Inviscid Flows

In computations of compressible flows the conservative form is most frequently

used, for reasons that we will discuss later. Furthermore, the viscosity is often
taken to vanish. The equations are then generally written as:
U E F G
+
+
+
=0
t
x
y
z

(19)

where

u
v
w
2
uv
uw
u
u + p

uv
vw
U = v ; E =
; F = v 2 + p ; G =
.

2
w
uw
vw
w + p
E
(E + p)u
(E + p)v
(E + p)w
(20)
If the gas is assumed to obey the perfect equation of state, then

P = RT,

e = cv T,

and

h = cp T.

(1.1)

Here R is the gas constant and = cp /cv . From these equations we find that
cv =

R
;
1

P = ( 1)e;

and

T =

( 1)e
.
R

(1.2)

1.2. SPECIAL CASES

d
dt
d
dt
d
dt

Z
V

I
u nds = 0

dv +
V

Z
udv =

f dv +

1
(e + u2 )dv =
2

I 


nT u(u n) ds

I
u f dv +

1
n (uT (e + u2 )u q)ds
2
S

Figure 1.1: The equations of fluid motion in integral form. The flux terms have
been moved under the same surface integral as the stress terms.

+ (u) = 0
t
u
= f + (T uu)
t



1
1
(e + u2 ) = (e + u2 )u Tu + q
t
2
2
Figure 1.2: The equations of fluid motion in conservative form.
D
+ u = 0
Dt

Du
= f + T
Dt

De
= T u q
Dt

Figure 1.3: The equations of fluid motion in convective form.

1.2.2

CHAPTER 1. THE EQUATIONS OF FLUID DYNAMICSDRAFT

Incompressible Flows

For an important class of flows the density of a material particle does not change
as it moves with the flow. The equation of state is therefore
D

=
+ u = 0
Dt
t

(21)

and the continuity equation reduces to

u = 0.

(22)

This states that the volume of any fluid element can not be changed and these
flows are therefore usually called incompressible flows. By expanding the divergence in (2) and using the equation of state we find that
I
u nds = 0
(23)
S

for incompressible flows. Notice that there is no restriction to constant density

flows here, the density of a material particle can vary from one particle to the
next one, but the density of each particle must stay constant. The density field
is found by solving the equation of state, eq. (6), but if the density is constant
everywhere that is not necessary.
The special case of incompressible fluids with constant density and viscosity
is of considerable importance, and we will devote much space to that. The
equations to be solved are then (22) and
u
p
+ u u =
+ f + 2 u
t

(24)

where = / is called the kinematic viscosity. This is the form usually seen
in text books, but we note that the nonlinear term can be expressed in several
forms:
1
u u = uu = u u u ( u).
(25)
2
For numerical predictions equations (22) and (24) are actually somewhat
awkward: We have evolution equations for each of the velocity component, and
one expression for the relation between the velocities but really no separate
equation for the pressure. This can be corrected by taking the divergence of the
momentum equation and using the incompressibility conditions to eliminate the
time derivative. This leads to
2 p = (u u) + f

(26)

which replaces the incompressibility equation. We note the special role played
by the pressure for incompressible fluid. Instead of being a thermodynamic function of (say) density and temperature, it is determined solely by the velocity
field. Indeed, it will take on whatever value is necessary to make the flow divergence free. It is sometimes convenientand we will use this extensivelyto

1.2. SPECIAL CASES

think of the pressure as projecting the velocity field into the space of incompressible functions. That is, we imagine the velocity first being predicted by (24)
without the pressure, then we adjust the pressure in such a way as to enforce
incompressibility and correct the velocities.
For incompressible fluids, where the density of a material particle is constant,
the term expressing work done by compressing the fluid is absent. The internal
heat generation due to viscous dissipation is frequently small, allowing us to
ignore . With these assumptions the energy equation reduces to
cv

DT
= kT
Dt

(27)

and is decoupled from the rest of the equations in the sense that while T depends
on the velocity field, the velocity field does not depend on the temperature.

1.2.3

Vorticity form of the incompressible equations

The equations for incompressible flows can be rewritten, and integrated for
special cases, by introducing two potentials such that the velocity is given by
u = +

(28)

This is generally referred to as Helmholtzs decomposition, or as the fundamental theorem of vector analysis (for a proof see, e.g. []) and it can be shown,
under fairly general conditions, satisfied in our case, that any vector field can
be decomposed in this way. To derive equations for these potentials we take
first the divergence of the definition (28) yielding
2 = u = 0

(29)

( ) = u =

(30)

and then the curl

Equation (30) may be simplified further if we note that it is really only the
velocity field that we are interested in. We may therefore add a function to be
as long as its curl remains the same. In particular, adding , where is a
scalar function, produces the same velocity field. We use this to set = 0.
The reason why this is allowable can be seen as follows: Suppose 6= 0. If
we add where 2 = then obviously (+) = = 0
and if we work with 0 = + , this potential gives the same velocity field
and is divergence free. Working with 0 (and dropping the prime) (30) becomes
2 =

(31)

( ) = ( ) 2 .

(32)

where we have used that

This transform to a divergence free vector potential is called a Gauge Transform.

CHAPTER 1. THE EQUATIONS OF FLUID DYNAMICSDRAFT

To obtain an equation for we take the curl of the Navier-Stokess equations

(24)
1

D
p
= ( )u + ( u) (
)+
(u + T u)
(33)
Dt

For incompressible, constant density flow, this equation reduces to

D
= ( )u + 2 .
(34)
Dt
When the flow is two-dimensional, both the vector stream function, and the
vorticity vector have only one non-zero component,
= (0, 0, ),

= (0, 0, )

(35)

where

v
u

.
x y
The right hand side of (9) is then identically zero, and we have to solve
=

D
= 2 ,
Dt

2 = .

(36)

(37)

The velocity is
u = k = k

(38)

or

v=
.
(39)
y
x
The steam function can be eliminated in favor of the velocity components
by taking the curl of equation (37). This results in an elliptic equation relating
the velocity components to the curl of the vorticity
u=

2 u = ,
or
2 u =

;
y

(40)

2 v =

(41)

in two dimension.
We have seen above that there are three basic form of the equations that
may be used as the basis of an numerical approximation. Which of these form
is best depends on the dimensions of the problem, the boundary conditions,
and what the programmer is used to. The number of equations is obviously an
important factor and the following table summarizes that.
Method
up
up

u
u

Dimensions
2
3
2
3
2
3

# of elliptic eqs.
1
1
1
3
2
3

# of advection eqs.
2
3
1
3
1
3

1.2. SPECIAL CASES

u
p
+ u u =
+ f + 2 u
t

2 p = (u u)
Primitive variables

+ ( ) u = 2
t
2 =
Streamfunction-vorticity variables

+ u u = 2
t
2 u =
Velocity-vorticity variables

Figure 1.4: Three forms of the incompressible Navier-Stokes equations

Although the number of equations does vary, depending on which formulation is used, their fundamental structure remains the same. In all cases we
have one or more advection-diffusion equation describing the evolution of certain
quantities with time and one or more elliptic equation, representing a quantity
that at each instant has equilibrated over the whole domain. The nature of the
boundary conditions does though change from one formulation to the next.